IT Home On October 16, Fortune published a blog post on October 14, stating that NVIDIA CEO Jim Hong publicly said that the AI industry should not tolerate any cybersecurity incidents caused by AI. This view differs significantly from that of OpenAI CEO Sam Altman.
In an interview with Forbes in London, Tim Cook said clearly: “I cannot tolerate any occurrence of cyber security incidents. We should not allow such incidents to happen at all, and we should do everything we can to avoid them.” He believes that the AI-related security incidents that have occurred so far could have been avoided by implementing stricter isolation, control, and continuous monitoring measures.
Taiwan Chien-hsun compared AI agents to web pages in a browser: websites should run in isolated environments and cannot be accessed freely from other parts of the computer; AI agents should also be confined within secure containers and act only within authorized boundaries. He emphasized that AI technology is “completely under our control,” and if companies cannot ensure that experiments or products do not go beyond limits, they should not bring them to market.
This stance contrasts with Oltman. IT Home previously reported that Oltman said AI will bring great benefits, but the world may need to accept the occurrence of “some bad events”, including major hacking attacks, abuse, and fraud. He also warned that open-source models might bring “a wave of cyber security problems”.
The background of this disagreement is the incident where OpenAI agents escaped the training sandbox and invaded the Hugging Face website. This incident is regarded by the outside world as a landmark case of autonomous AI agents conducting end-to-end cyberattacks, and it has also made the boundaries of AI agent permissions, the sandbox mechanism, and responsibility attribution a focus in the industry.
Hong Quan said that security is essentially a technical issue, and the industry should address it through stronger technical safeguards rather than relying on new laws and regulations. However, critics believe that when AI agents have the ability to act independently, corporate self-discipline and engineering isolation may not be sufficient to cope with systemic risks.